How to introduce yourself in an AI interview: building a 2-minute pitch

Why this question decides the rest of the interview

In an AI interview, the person opposite is trying to place three things quickly: the kind of problems you can actually handle, the level of responsibility you have genuinely held, and the distance between your experience and their need. A chronological answer — degree, first job, second job — answers none of them. It forces the interviewer to do the sorting, and you lose the two most attentive minutes of the conversation.

So give them a frame instead, in this order: who I am professionally today → what I can solve → proof → what I am looking for. Everything else in the interview will slot into that frame, and it lets you steer the follow-up questions towards ground you know well.

The structure of a 2-minute pitch

Four parts, in this order. The timings are orders of magnitude, not a stopwatch: what matters is the balance between them. A pitch that spends half its time on education is unbalanced whatever its length.

1. 1

   Your current positioning

   ~20 s

   One sentence about who you are today, not who you have been. “I'm an AI Engineer working on LLM applications running in production in a regulated environment.” No degree list, no history. If you are transitioning, say so plainly rather than hiding it — it reads far stronger than a vague position.

2. 2

   Your specialisation

   ~30 s

   Which problems can you genuinely handle, and under what conditions? Document retrieval over messy corpora, industrialising existing models, reducing inference cost, hardening a system already in production… Name two or three families of problems, not ten technologies. This is the part that separates you from other candidates with the same job title.

3. 3

   One representative project

   ~50 s

   A single project, told in four moves: problem → constraints → decisions → outcome. Pick the project that matches what the company needs, not the most impressive one on your CV. This is your proof — without it, the first two parts remain claims.

4. 4

   What you are looking for

   ~20 s

   Connect your experience to the role explicitly. “What interests me here is the production side at real volume, because that is exactly what I've been building for the past two years.” That sentence gives the interviewer a reason to continue and tells them which questions to ask next.

Two minutes is short. Prepare the pitch, but do not learn it word for word: a recited answer is obvious within seconds and makes the rest of the conversation stiffer. Remember the four anchors and phrase them freely.

A concrete example: AI Engineer profile

Same background, two ways of presenting it. The first is very common; it is not wrong, it is simply unusable for the person listening.

How to present an AI project

Listing “Python, LangChain, OpenAI, RAG, AWS” says nothing about your level: anyone can list the same stack. What places you is the chain of decisions. Tell the project in this order: problem → constraints → choices → trade-offs → evaluation → production → outcome.

In AI roles, technical teams listen most closely to the following — because these are what separate a demo from a system that holds:

• Data: where it came from, what state it was in, what had to happen before anything else.
• Evaluation: how you knew the system was improving, and who defined what a good answer was.
• Limits: what the system cannot do, and how that limit is handled in the product.
• Cost, when it is a criterion: cost per query, latency, model size versus quality.
• Reliability: failure behaviour, guardrails, what gets escalated to a human.
• Production: deployment, monitoring, what broke and what you learned from it.
• Your decisions: what you settled yourself, versus what the team decided or the context imposed.

One last point, often decisive: own the collective part. Saying “the data team owned the ingestion pipeline, I was on evaluation and serving” is more credible than claiming everything, and it has never weakened a serious application.

Adapting the pitch to your profile

Junior

Do not compensate for limited experience with a longer tool list. Pick one project — academic, personal or from an internship — and treat it exactly like professional work: problem, constraints, decisions, what you measured, what failed. One project told with that rigour beats five skimmed ones, and it demonstrates the very skill being assessed.

Mid-level / senior

The subject is no longer technique alone but scope: architecture decisions, cost/quality trade-offs, what you refused to build and why, mentoring or technical review. One project is enough, chosen for its closeness to the company's need. Avoid the career panorama — it dilutes your positioning.

Freelance / consultant

Introduce yourself through the problems you solve and how you work, not through a client list. Say which contexts you are effective in, what you leave behind at the end of an engagement — documentation, handover, team autonomy — and at which stage of a project you usually step in.

Moving into AI from another field

Name the transition in one sentence, without apologising, then move to what genuinely transfers: engineering rigour, domain knowledge, production experience, data handling. Anchor it in one concrete AI piece of work, however modest, and be precise about what you have not mastered yet. That clarity reassures far more than a layer of vocabulary.

Common mistakes

• Reciting your CV chronologically: the interviewer has already read it.
• Chaining technologies without saying what you did with them.
• Talking for five minutes: past two or three, attention drops and you lose the thread.
• Describing a project without its context: with no problem, the outcome has no value.
• Showing no trade-off: a project without constraints sounds like an exercise.
• Overstating your level: one follow-up question is enough to reveal it.
• Using AI terms you cannot explain in plain words.
• Never connecting your background to the company's need — the one thing they must remember.

• Interviews
• AI Engineering
• Evaluation